针对非理想拍摄条件下得到的图片质量较差导致标定精度下降的问题,提出一种稳健的标定方法.首先构建出一种新的标定参照物,并且采用了后测量的策略确定标定物的尺寸,降低了测量中的相对误差.其次,引入RGB空间信息,辅助进行特征检测,提高了检测的自动化程度,消除了特征的误匹配.实验结果显示,标定的精度与稳健性相比棋盘模板方法有所提高,并且具有极小的反投影误差.%A robust calibration approach is proposed. Most of the classical experiments of calibration are conducted with ideal calibration images, but the non-ideal images are frequently used in practice. In order to overcome the degeneration of traditional calibration methods, a novel compact setup of calibration object is proposed. A post-measure strategy is introduced to measure the calibration objects which can enhance both the noise tolerance and precision of measurement. The strategy" can also decrease the relative error of feature detection theoretically. The Color Information Aided (CIA) and ellipse detection are employed to automatically detect the features. Consequently, the method can guarantee the correspondence between image coordinates and world coordinates without concerning about the global rotation of features. The experiments are performed with both the proposal and chessboard based method which is used frequently.
展开▼
